1. Michael Ernst UW CSE 140 Winter 2013
Sorting
Michael Ernst
UW CSE 140
Winter 2013

2. Sorting
hamlet = "to be or not to be that is the question whether tis nobler in the mind to suffer".split()
print "hamlet:", hamlet
print "sorted(hamlet):", sorted(hamlet)
print "hamlet.sort():", hamlet.sort()
Lists are mutable – they can be changed
including by functions

3. Customizing the sort order
Goal: sort a list of names by last name
names = ["Isaac Newton", "Albert Einstein", "Niels Bohr", "Marie Curie", "Charles Darwin", "Louis Pasteur", "Galileo Galilei", "Margaret Mead"]
print "names:", names
This does not work:
print "sorted(names):", sorted(names)
When sorting, how should we compare these names?
"Niels Bohr"
"Charles Darwin"
Goodall, Jane, English Mead, Margaret,  Marie Curie

4. Sort key
A sort key is a different value that you use to sort a list, instead of the actual values in the list
def last_name(str):
return str.split(" ")[1]
print 'last_name("Isaac Newton"):', last_name("Isaac Newton")
Two ways to use a sort key:
Create a new list containing the sort key, and then sort it
Pass a key function to the sorted function

5. 1. Use a sort key to create a new list
Create a different list that contains the sort key, sort it, then extract the relevant part
# keyed_names is a list of [lastname, fullname] lists (tuples would be better!)
keyed_names = []
for name in names:
keyed.names.append([last_name(name), name])
# simpler: keyed_names = [[last_name(name), name] for name in names]
print "keyed_names:", keyed_names
print "sorted(keyed_names):", sorted(keyed_names)
print "sorted(keyed_names, reverse = True):"
print sorted(keyed_names, reverse = True)
(This works because Python compares two elements that are lists elementwise.)
sorted_keyed_names = sorted(keyed_names, reverse = True)
sorted_names = []
for keyed_name in sorted_keyed_names:
sorted_names.append(keyed_name[1])
# simpler: sorted_names = [keyed_name[1] for keyed_name in sorted_keyed_names]
print "sorted_names:", sorted_names

6. Digression: Lexicographic Order
Aaron
Andrew
Angie
[1, 9, 9]
[2, 1]
[3]
with
withhold
withholding
[1]
[1,1]
[1,1,1]
Each of these lists is in lexicographic order
Able
Charlie
baker
delta

7. 2. Use a sort key as the key argument
Supply the key argument to the sorted function or the sort function
print "sorted(names, key = last_name):"
print sorted(names, key = last_name)
print "sorted(names, key = last_name, reverse = True):"
print sorted(names, key = last_name, reverse = True)
print sorted(names, key = len)
def last_name_len(name):
return len(last_name(name))
print sorted(names, key = last_name_len)

8. itemgetter is a function that returns a function
import operator operator.itemgetter(2, 7, 9, 10) operator.itemgetter(2, 7, 9, 10)("dumbstricken") operator.itemgetter(2, 5, 7, 9)("homesickness") operator.itemgetter(2, 7, 9, 10)("pumpernickel") operator.itemgetter(2, 3, 6, 7)("seminaked") operator.itemgetter(1, 2, 4, 5)("smirker") operator.itemgetter(9, 7, 6, 1)("beatnikism") operator.itemgetter(14, 13, 5, 1)("Gedankenexperiment") operator.itemgetter(12, 10, 9, 5)("mountebankism")

9. Using itemgetter Sort the list by name: Sort the list by score
from operator import itemgetter
student_score = ('Robert', 8)
itemgetter(0)(student_score)  “Robert”
itemgetter(1)(student_score)  8
student_scores =
[('Robert', 8), ('Alice', 9), ('Tina', 7)]
Sort the list by name:
sorted(student_scores, key=itemgetter(0) )
Sort the list by score
sorted(student_scores, key=itemgetter(1) )

10. Sorting based on two criteria
Two approaches:
Use an itemgetter with two arguments
Sort twice (most important sort last)
student_scores = [('Robert', 8), ('Alice', 9), ('Tina', 10), ('James', 8)]
Goal: sort based on score; within score, by name
sorted(student_scores, key=itemgetter(1,0) )
result = sorted(student_scores, key=itemgetter(0) ) result = sorted(result, key=itemgetter(1) )

11. More sorting based on two criteria
If you want to sort different criteria in different directions, you must use multiple calls to sort or sorted
student_scores = [('Robert', 8), ('Alice', 9), ('Tina', 10), ('James', 8)]
Goal: sort score from highest to lowest; within score, sort name alphabetically (= lowest to highest)
result = sorted(student_scores, key=itemgetter(0) ) result = sorted(result, key=itemgetter(1), reverse=True)

12. Sorting: strings vs. numbers
Sorting the powers of 5:
>>> sorted([125, 5, 3125, 625, 25])
[5, 25, 125, 625, 3125]
>>> sorted(["125", "5", "3125", "625", "25"])
['125', '25', '3125', '5', '625']

13. This “default” pattern is so common, there is a special method for it.
Top k with a dictionary
>>> uniquewords = { ‘lol’:45, ‘omg’:23, ‘know’:12 }
>>> uniquewords[‘omg’]
def top10(doc):
“““Return a list of the top 10 most frequent words.”””
uniquewords = {}
for word in doc:
if uniquewords.has_key(‘omg’):
uniquewords[‘omg’] = uniquewords[‘omg’] + 1
else:
uniquewords[‘omg’] = 1
# now search for top 10 most frequent words
bycount = [(pair[1], pair[0]) for pair in uniquewords.items()]
bycount.sort()
return bycount[0:10]
This “default” pattern is so common, there is a special method for it.

14. This “default” pattern is so common, there is a special method for it.
Top k with a dictionary
>>> uniquewords = { ‘lol’:45, ‘omg’:23, ‘know’:12 }
>>> uniquewords[‘omg’]
def top10(doc):
“““Return a list of the top 10 most frequent words.”””
uniquewords = {}
for word in doc:
uniquewords[‘omg’] = uniquewords.set_default(‘omg’, 0) + 1
# now search for top 10 most frequent words
bycount = [(pair[1], pair[0]) for pair in uniquewords.items()]
bycount.sort()
return bycount[0:10]
This “default” pattern is so common, there is a special method for it.